GitHub announced Wednesday that its Copilot coding agent now includes web browser functionality, allowing the artificial intelligence tool to interact with web applications while making code changes.
The new feature, currently in public preview, integrates Playwright’s browser automation technology through the Model Context Protocol server system. The update affects all paid Copilot subscribers using the coding agent service.
“This allows Copilot to reproduce bugs and validate its work,” GitHub said in a blog post announcing the feature.
The coding agent operates in a separate development environment when users delegate programming tasks. With browser access enabled by default, the AI can now test web applications directly, take screenshots of its work and include visual documentation in pull requests.
GitHub’s Copilot coding agent launched earlier this year as a more autonomous version of the company’s AI programming assistant. The tool can work independently on coding tasks rather than requiring constant user input.
The browser functionality represents a significant expansion of Copilot’s capabilities, moving beyond code generation to include testing and validation of web-based applications. Users can also configure additional tools through the Model Context Protocol system.
Access to the coding agent requires a paid Copilot subscription, either Copilot Business or Copilot Enterprise. Organizations using enterprise versions must have administrators enable the feature before employees can use it.
GitHub, owned by Microsoft Corp., has positioned its AI coding tools as essential productivity enhancers for software developers. The company has not disclosed specific user numbers for the coding agent feature.
The web browser update comes as competition intensifies in the AI-assisted programming market, with companies like Amazon, Google and OpenAI offering similar developer tools.
GitHub provides documentation and best practices for users looking to implement the coding agent in their development workflows.